|
Yeah IIRC those Palit cards are the absolute best for noise and thermals but they're not available everywhere.
|
#
?
Nov 9, 2017 22:15
|
|
|
|
| # ? May 15, 2024 07:49 |
|
|
GRINDCORE MEGGIDO posted:The Palit Gamerock 1070 must own thermally. The 1080 does. I got a Palit Super JetStream 1070, because it was the only thing available for semi-reasonable €€€ back then. The only time I've ever actually heard it was when it was pulsing the fans on the edge of the 0dB mode. They fixed that along with the BIOS update for the wonky memory stuff and now I just never hear it. It never goes above 40% fan speed.
|
|
#
?
Nov 10, 2017 00:03
|
|
|
Generic Monk posted:is there an actual writeup on what fast sync actually does under the hood? when i had a gaming pc i forced it on for everything and there really was a palpable improvement in input latency and general... smoothness, but pretty much every mention i saw of it was just quoting the nvidia line of it being a 'supercharged version of vsync' or some poo poo like that. i totally expected it to be snake-oil from that but it was actually really great, i've just got no idea what it actually does https://youtu.be/L07t_mY2LEU You need to match your refresh rate in FPS and it is wonky on some games. I use it in Overwatch and Destiny only, because to get the most of it in Windows 8/10 you have to run the game full screen (or else you’re getting regular Vsync from the window composer), and I like borderless too much in games where quick draw headshots aren’t a gameplay mechanic.
|
|
#
?
Nov 10, 2017 01:38
|
|
|
my gpu fan is ramping up and down from 0 to 100. The fan curve should have it at 100 and its not dropping the temp anywhere near below that threshold so it shouldnt going up and down according to the temp and fan curve. Does it indicate a fan going bad or power delivery to fan failing or maybe conflicting drivers both trying to control the fan?
|
|
#
?
Nov 10, 2017 03:14
|
|
|
Pascal or earlier? With Pascal the only way to set a fan curve is with a hack (continually resetting a fixed fan speed). I'd try setting it to a fixed speed first and see how it handles that. When I saw the fan control options available with Pascal I expected them to break at any moment. My dumb solution was using externally controlled fans to bypass all that, works, but the card is obviously modified. If it's something earlier, maybe try uninstalling and reinstalling Afterburner? Maybe reflashing the bios because something got messed up with a new driver update? I don't think they've broke older cards fan speed control, my 980 still is using the custom fan curve I gave it no problem. Anything else, maybe the fan motor or card PWM fan controller is failing.
|
|
#
?
Nov 10, 2017 03:32
|
|
|
Is this just AMD reassuring their shareholders, or is AMD a little shook by Raja bailing to run Intel's discrete GPU program?quote:AMD sent the following statement when asked for comment on Intel's new hire:
|
|
#
?
Nov 10, 2017 07:42
|
|
|
I mean after Zenimax v Oculus these kinds of moves are going to always be looked at as an opportunity to quickly make some bucks.
|
|
#
?
Nov 10, 2017 10:01
|
|
|
Raja's comment on how fantastically amazing the opportunity to work at Intel was seemed very much like he's sucking some balls.
|
|
#
?
Nov 10, 2017 10:39
|
|
|
Here's an interesting HardOCP article from May 2016 with background info on the whole deal. https://www.hardocp.com/article/2016/05/27/from_ati_to_amd_back_journey_in_futility
|
|
#
?
Nov 10, 2017 11:34
|
|
|
The fact that nothing absolutely insane has happened in GPU would this morning is a little disappointing given the rest of the week.
|
|
#
?
Nov 10, 2017 16:08
|
|
|
Samsung just announced that they have 16Gbps GDDR6 in the pipe: https://twitter.com/VideoCardz/status/929020970945404928 If it's ready in time for consumer Volta then GV104 could hit 512GB/sec on 256-bit and GV102 could hit 768GB/sec on 384-bit 
|
|
#
?
Nov 10, 2017 17:53
|
|
|
NewFatMike posted:The fact that nothing absolutely insane has happened in GPU would this morning is a little disappointing given the rest of the week. I'm sure Glaze3d will be hitting retail later this afternoon.
|
|
#
?
Nov 10, 2017 18:23
|
|
|
Palladium posted:Then there are those put max brightness of a million suns on everything and then are completely surprised that they fail shortly after the warranty period. If you have a TV that has full array local dimming (or OLED) and HDR, you'll most likely get a better picture with max brightness.
|
|
#
?
Nov 10, 2017 18:36
|
|
|
repiv posted:Samsung just announced that they have 16Gbps GDDR6 in the pipe: Nice, very nice. 
|
|
#
?
Nov 10, 2017 18:57
|
|
|
NewFatMike posted:The fact that nothing absolutely insane has happened in GPU would this morning is a little disappointing given the rest of the week. repiv posted:Samsung just announced that they have 16Gbps GDDR6 in the pipe: Eh, I guess this counts. I was hoping for like the reveal of a murder conspiracy or something.
|
|
#
?
Nov 10, 2017 20:09
|
|
|
We got silicon. PCPer/Chiphell says that's a Hades Canyon NUC board. https://www.pcper.com/news/General-Tech/Rumor-Hades-Canyon-NUC-AMD-Graphics-Spotte Jesus, look at how many power phases that thing has.
|
|
#
?
Nov 10, 2017 22:13
|
|
|
SwissArmyDruid posted:We got silicon. PCPer/Chiphell says that's a Hades Canyon NUC board. https://youtu.be/Av8sn7BXLxE I’ll be real interested to play with these as well. Hopefully there are unlocked multipliers on both the CPU and GPU, it would be neat to compare to Raven Ridge APUs.
|
|
#
?
Nov 10, 2017 22:23
|
|
|
NewFatMike posted:https://youtu.be/Av8sn7BXLxE Don't think Raven Ridge is going to come out looking pretty.
|
|
#
?
Nov 10, 2017 22:29
|
|
|
I like how "ChipHell" gets the Hades Canyon chip, and there's Devil's Canyon too. Has Intel been hailing Satan all this time? Explains a lot of things...
|
|
#
?
Nov 10, 2017 22:31
|
|
|
SwissArmyDruid posted:Jesus, look at how many power phases that thing has. 6 for the CPU, 5 for the GPU and 1 for the HBM. It's a lot, but doesn't seem unreasonable. I'm just guessing based on their physical location. It could be 12 all feeding a unified voltage to everything.
|
|
#
?
Nov 10, 2017 22:52
|
|
|
craig588 posted:6 for the CPU, 5 for the GPU and 1 for the HBM. It's a lot, but doesn't seem unreasonable. I don't think there can be a unified voltage because the CPU and GPU probably have to operate at their own independent voltages, right?
|
|
#
?
Nov 10, 2017 22:57
|
|
|
Yeah, that's what I assume, but who knows what Intel cooked up. I wouldn't rule it out as impossible, though it is unlikely.
|
|
#
?
Nov 10, 2017 23:12
|
|
|
Man, at what point will they shift to a daughter-board riser card with just all the power phases on it?
|
|
#
?
Nov 10, 2017 23:55
|
|
|
NewFatMike posted:The fact that nothing absolutely insane has happened in GPU would this morning is a little disappointing given the rest of the week. This Intel thing didn’t happen overnight. It took some time and while it looks to us like everyone is backstabbing each other, the people who actually decide things knew what was going to happen before the dominos fell.
|
|
#
?
Nov 10, 2017 23:58
|
|
|
Methylethylaldehyde posted:Man, at what point will they shift to a daughter-board riser card with just all the power phases on it? https://www.evga.com/products/product.aspx?pn=100-UV-0600-BR For the extreme overclocking crowd we're already there.
|
|
#
?
Nov 11, 2017 00:07
|
|
|
That chip is so going to end up in a new Mac Mini 
|
|
#
?
Nov 11, 2017 00:35
|
|
|
GRINDCORE MEGGIDO posted:Don't think Raven Ridge is going to come out looking pretty. It's more with what's unlocked on each one. If Ryzen design decisions carry over, I'd expect there to be some room to mess with voltages and clocks on the CPU and GPU. Craptacular! posted:This Intel thing didn’t happen overnight. It took some time and while it looks to us like everyone is backstabbing each other, the people who actually decide things knew what was going to happen before the dominos fell. I'm going to assume you're not being condescending, but yeah custom silicon has considerable lead time. That with the Raja resigning and the next day announcing he's working for Intel is the other part of this buck wild week.
|
|
#
?
Nov 11, 2017 02:22
|
|
|
SwissArmyDruid posted:We got silicon. PCPer/Chiphell says that's a Hades Canyon NUC board. That thing is going to come with a power brick almost as big as the unit itself.
|
|
#
?
Nov 11, 2017 03:23
|
|
|
Krailor posted:That thing is going to come with a power brick almost as big as the unit itself. xbox has been doing it for awhile and people seem pretty over it
|
|
#
?
Nov 11, 2017 03:50
|
|
|
Krailor posted:That thing is going to come with a power brick almost as big as the unit itself. It's better to have a seperate power brick as it removes a heat producing part from the enclosure and they're passively cooled. Plus I'd rather have a brick on the floor rather than that much more volume on my desk.
|
|
#
?
Nov 11, 2017 03:54
|
|
|
NewFatMike posted:I'm going to assume you're not being condescending, but yeah custom silicon has considerable lead time. Kind of. I’ve been reading Reddit posts that treat the tech industry like it’s some sort of sports team competition or edgy drama, and I’ve had family who work in the field of hardware production and manufacturing all my life. I know this poo poo must have been building for months and appeared to have happened in 48 hours because it can at last be disclosed to shareholders.
|
|
#
?
Nov 11, 2017 03:59
|
|
|
Craptacular! posted:Kind of. I’ve been reading Reddit posts that treat the tech industry like it’s some sort of sports team competition or edgy drama, and I’ve had family who work in the field of hardware production and manufacturing all my life. I know this poo poo must have been building for months and appeared to have happened in 48 hours because it can at last be disclosed to shareholders. Oh yeah, I mean I can't expect you to know, but I do a lot of manufacturing stuff, and even production runs on barely machined metal takes like 2 months to ramp up. Custom silicon tapeouts must be enormous in comparison. I don't envy the Reddit posts, I'm pretty glad not to be on there anymore.
|
|
#
?
Nov 11, 2017 04:06
|
|
|
Fauxtool posted:xbox has been doing it for awhile and people seem pretty over it Xbox's PSU has been internal since the S years ago
|
|
#
?
Nov 11, 2017 04:35
|
|
|
Krailor posted:That thing is going to come with a power brick almost as big as the unit itself. THAT is what I was thinking when I saw that many power phases. Here's hoping Intel doesn't ship with one of those stupid loving ones with the cable from the brick to the wall ungodly long, and the cable from the brick to the device being too goddamn short.
|
|
#
?
Nov 11, 2017 05:29
|
|
|
Happy_Misanthrope posted:Xbox's PSU has been internal since the S years ago 1 year ago was the xbox one s. It was happening for over a decade. The point stands Fauxtool fucked around with this message at 06:17 on Nov 11, 2017 |
|
#
?
Nov 11, 2017 06:12
|
|
|
http://techreport.com/news/32821/samsung-teases-higher-than-expected-speeds-from-its-gddr6-ram#metal GDDR6 already hitting 16gbps, poor Vegas.
|
|
#
?
Nov 11, 2017 09:16
|
|
|
I've been idly thinking about a 4x 1080 Ti build for a hobby ML project (the benchmarks for the frameworks I use point to linear scaling with # of GPUs) and had some questions for you guys - 1) is nvidia likely to come out with a 1180/Ti equivalent in the next half a year? 2) what CPU platform would be a good fit? I'm happy with my current Ryzen 1700 but a quick Google seems to be suggesting that if I'm getting 4x GPUs, I should be getting a i9 7XXX series or TR series because the Ryzen and i7 mobos don't seem build for that many GPUs. is that right? 3) i was thinking of the Zotac 1080 Ti mini from a space standpoint. is there a specific model/make I should be looking at? 4) are there any other misc. considerations I should be aware of? thanks!
|
|
#
?
Nov 11, 2017 12:23
|
|
|
shrike82 posted:I've been idly thinking about a 4x 1080 Ti build for a hobby ML project (the benchmarks for the frameworks I use point to linear scaling with # of GPUs) and had some questions for you guys - 1. Yes, the Volta x80 is coming in spring, but based on past trends the 1080 ti will only be a bit slower than it and if you can use them now it's probably not worth waiting 2. If you think you'll be bus limited a Threadripper or Skylake-X is the only option to get tons of PCI-E lanes. If you're not going to be hitting the bus hard even a AMD 1300 or Intel 8100 could be fine, if you don't need CPU performance. 3. Zotacs are really loud because of their crappy fans, but there's nothing functionally wrong with them besides the noise 4. Games will not use 4X Nvidia GPUs anymore, Nvidia capped the scaling at 2 so there's no side benefit if you thought it might work double as a compute and gaming PC.
|
|
#
?
Nov 11, 2017 13:23
|
|
|
Fauxtool posted:1 year ago was the xbox one s. It was happening for over a decade. The point stands yeah i don't think it's too controversial to say that the xbox up to the introduction of the s was not exactly a bastion of great hardware design. honestly in the vast majority of cases if your thing has a power brick and isn't hugely thinness-constrained i'm going to assume that you didn't really give a poo poo about your hardware design. which is fine - maybe you had other priorities - but yeah shrike82 posted:I've been idly thinking about a 4x 1080 Ti build for a hobby ML project (the benchmarks for the frameworks I use point to linear scaling with # of GPUs) and had some questions for you guys - why are you looking at a card marketed as 'mini' from a 'space saving standpoint' when you're planning to run four of the cards? you're going to need a gigantic case anyway, the length of the card is going to be the least of your problems Generic Monk fucked around with this message at 14:36 on Nov 11, 2017 |
|
#
?
Nov 11, 2017 14:33
|
|
|
|
| # ? May 15, 2024 07:49 |
|
|
shrike82 posted:I've been idly thinking about a 4x 1080 Ti build for a hobby ML project (the benchmarks for the frameworks I use point to linear scaling with # of GPUs) and had some questions for you guys - Another option for a bunch of pcie lanes would be to build an x99 system. It's a little older but plenty of places are still selling stock and it will be much cheaper than either a Threadripper or x299 system. Also if you're going to get 4x GPUs you want to get ones with blower style coolers; they're made for sitting close together in SLI. The open air coolers are great if there's space between cards but they start choking when the cards are right next to each other.
|
|
#
?
Nov 11, 2017 14:50
|
|